A study of data visualisation methods for reconstructing the symbolic transmission of Aboriginal oral traditions

In a narrow sense, “indigenous” is used to describe only minorities living in a country or region dominated by an alien majority. In a broader sense, the United Nations Sub-Commission on Prevention of Discrimination and Protection of Minorities describes Indigenous communities, peoples and nations as “those who, in historical continuity with the societies that developed in their territories prior to occupation or colonization, regard themselves as distinct from the other groups of people who now dominate the territory of the State or part of the State”.

According to these definitions, Indigenous communities can be understood as marginalized people and minorities that were historically colonized but still have their original territories and social systems. Their cultural traditions are thus distinct from what the public generally refers to as traditional or folk culture. The main indigenous communities at present include the Indians of the Americas, the Maori of Oceania, the Sami of Northern Europe, and the Inuit near the Arctic Circle.

After years of research and investigation, the World Intellectual Property Organization (WIPO) has established a series of legal terms for the definition and protection of traditional tribal folk culture and art, which is a major theoretical achievement in the international intellectual property legal system. The three main terms used are “traditional knowledge”, “traditional cultural expressions” (TCEs) and “genetic resources”. Among them, “traditional cultural expressions” is also known as “expressions of folk culture and art”. It is defined by the World Intellectual Property Organization (WIPO) as “the product of creations reflecting the characteristics of a traditional cultural heritage developed and held by a community or ethnic group in a country, or by certain individuals representing the traditional cultural expectations of that group”, and includes such forms as oral expressions, musical expressions, and behavioral expressions. These terms are now widely used in the international legal context of intellectual property.

From the standpoint of semiotics, cultural transmission is based on symbols as the most basic means of communication and exchange. Among them, language is a symbol that is commonly agreed upon and shared by members of the nation. Language stores the memory of the nation and unites the will and centripetal force of the nation. Language serves as the record of national development and progress, the most significant cultural component of a nation, and the expression of national identity. Those oral creations widely circulated in folk culture can reflect the cultural characteristics of a nation from different aspects. Cultural inheritance can be divided into language inheritance, behavioral inheritance, artifact inheritance, psychological inheritance, and other forms of inheritance according to the form of cultural composition. One of the major signs of difference among nationalities is language. Oral archives, which transmit national culture through language, have naturally become symbolic tools.

Oral archives are responsible for producing and reproducing national culture. As a means of preserving national culture, oral archives are an important witness to national culture and a means of accumulating and disseminating history and culture. For example, the Guoyu and Zuozhuan of the pre-Qin period were written on silk and simplicity through the process of oral transmission, and were continuously processed and embellished. This type of oral archives has become a part of cultural accumulation, cultural inheritance and cultural protection that cannot be ignored.

With the progress and popularity of science and technology, audio-visual and other new media forms of oral archives have become a new force in the inheritance of national culture. The current oral archives, due to their flexible recording methods and diverse means of expression, have advanced cultural transmission from a purely traditional narrative to an advanced form that combines text and digital images. The methods of human expression have been enhanced by the vividness of its images and the operation of the original cultural transmission has been altered. Therefore, oral archives have provided a new platform for the inheritance and development of the cultural symbols of the indigenous peoples.

Oral archives are the source of precipitation and cohesion for the national psyche and national spirit. Oral archives are used to enhance traditional national cultural resources that rely on human memory. These national cultures, whether recorded in words or images, often express a strong national sentiment. The contents of oral archives include historical oral archives, literary oral archives, religious oral archives, ethical oral archives, and folklore oral archives. Many of them embody the psychological consciousness, values, and even emotional tendencies of a people, and together they contribute to the formation of a sense of national identity and cohesion.

Where, A denotes the source image. B denotes the structure of the convolution kernel.

Corresponding to expansion, erosion is the operation opposite to thresholding. Its mathematical definition equation is: 14A▯B={ x,y|B(x,y)⊆A }

If there is noise in the image, the expansion will increase the noise. Define the difference between the original image and the result of image erosion as the bright spots in the image that have been eroded by defining this P₁ difference as follows: 15P1=I−E(I)

Where I denotes the source image and E(I) denotes the result of the image after the corrosion operation. Similarly, the difference p₂ is defined as follows: 16P2=D(I)−I

Where, D(I) represents the result of the image after the expansion operation. The result P after enhancement is as follows: 17P=I+αP1−βP2

In the formula, the larger the values of α and β, the larger the contrast of the image, but the larger the contrast of the image is not better. The local maxima and local minima of the image computed by the expansion and erosion operations respectively. Therefore, the image bias variance is defined as follows: 18V(x)=1n∑i=1n(xi−e)2

Where x_i denotes the pixel value at point i: e is the local extreme value of the image, when solved α, e denotes the local extreme value, when solved β, e denotes the local extreme value, then α and β are defined as follows: 19α=AeVe,β=AdVd

Where A_c indicates the number of highlights, A_c increases indicating an increase in the bright areas of the image. A_d represents the number of dark areas, and A_d increase indicates the increase of dark parts of the image. The final image enhancement formula is as follows: 20{ P=I+α′(I−E(I))−β′(D(I)−I)α′=Aemin(top,max(floor,Ve))β′=Admin(top,max(floor,Vd))

Due to the high brightness of the spot in the laser array, it is mainly needed to attenuate the influence of the blank jumping line without additional enhancement, so the value of A_d should not be too large, otherwise it will make the luminance of the interference information enhancement, so the value of A_d is taken to be 0. The value of A_c is taken to be 30, in order to attenuate the influence of the interference information of the blanking jumping line.

For the input image firstly needs to be segmented into a collection of binary images under different thresholds by a series of successive thresholds to extract the connectivity domain of the binary images under different thresholds, the following equation shows the principle of dual threshold method segmentation. That is: 21g(x,y)={ 1,f(x,y)∈[ t1,t2 ]0,f(x,y)∈[ t1,t2 )∪( t2,255 ]

Where g(x, y) is the grey value of the input image at point p(x, y) after threshold segmentation, f(x, y) is the grey value of the input image at point p(x, y), t₁ is the low threshold, and t₂ is the high threshold.

In the field of image processing, the grey value of an image can be regarded as a two-dimensional density distribution function, then the characteristics of the image connectivity domain can be described by the moments, and the image moments are defined as follows: 22mp,q=∑i=1NI(xi,yi)xpyq

Where m_p,q represents the sum of all pixels in the image, where the pixel value I(x_i, y_i) of each pixel x, y is multiplied by a factor x^py^p. When denoting the zero-order moment m₀₀, this factor is equal to 1, i.e., m₀₀ denotes the region of the image where all the pixel values are non-zero, and hence the zero-order moment of the image can represent the area of the connected domain. Then the centre of mass coordinate (x¯,y¯) of the connected domain can be represented by this expected value: 23{ x¯=m10/m00y¯=m01/m00

Assuming that the connected domain is longer in one direction and shorter in its orthogonal direction, the longer direction of the connected domain can indicate the direction in which the domain is orientated. The moment of inertia E of the connected domain is defined as follows: 24E=∑i=1NI(xi,yi)ri2

Where r_i denotes the distance from the i nd pixel point on the connected domain to the rotary axis. From the parallel axis theorem, the rotary axis of minimum rotational inertia crosses the centre of mass (x¯,y¯) . Let the angle between the rotary axis and the x-axis be θ, then the linear equation of the rotary axis is: 25(x−x¯)sinθ−(y−y¯)cosθ=0

Then, the distance r_i from the i st pixel point on the connected domain to the rotary axis can be expressed as: 26ri=(xi−x¯)sinθ−(yi−y¯)cosθ

Then the moment of inertia E can be expressed as: 27E=∑i=0NI(xi,yi)[ (xi−x¯)sinθ−(yi−y¯)cosθ ]2

Where the value of the image centre distance μ_p,q is defined as follows: 28μp.q=∑i=1NI(xi,yi)(x−x¯)p(y−y¯)q

Then, substituting Eq. (27) into Eq. (28) gives: 29E=μ20sin2θ−μ11sinθcosθ+μ12cos2θ

Therefore, the problem for solving the minimum rotational inertia of the connected domain can be regarded as a quadratic optimisation problem with constraints, and its optimisation model is: 30{ Finds=(sinθ,cosθ)TminE=sTAss.t.1−sTIs=0

Where A=[ μ20μ11μ11μ02 ] , I are second order unit matrices. This quadratic optimisation problem with constraints can be solved by the Lagrange multiplier method by first constructing the Lagrange equation with multipliers as: 31L=sTAs−λ(1−sTIs)

The partial derivative of L with respect to s can be obtained by making 1 zero: 32∂L∂s=2As−2λIs=0

The above equation reduces to: 33As=λs

Substituting Eq. gives the moment of inertia E as: 34E=sTAs=λsTs=λ

Where λ is the eigenvalue of matrix A. Therefore, the solution of the rotational inertia of the connected domain is equivalent to solving the eigenvalue of the matrix A, where the eigenvector of the matrix A is the direction of the rotational inertia of the connected domain.

In order to achieve the calibration of the laser projection system, it is also necessary to match the light spots in the laser dot matrix collected by the camera with the control points of the laser dot matrix in the laser projection plane. In this paper, image matching of the laser spot array is carried out using the laser spot topology. Furthermore, the spot topology information can be used to further filter out any interference information that remains.

To verify the effectiveness of the design method, the oral translator speech recognition platform is constructed. The hardware construction of the oral interpreter speech recognition system is carried out through end-to-end speech feature analysis and the network-centric storage model structure analysis method.

On the basis of the above experimental platform, the performance of this paper’s method in aboriginal speech recognition is verified through simulation experiments. The transmission delay should not be greater than 500 ms, the input signal-to-noise ratio of the speech array sensor is -22 dB, the frequency deviation is 0~1500 kHz, the number of snapshots is 1500, the length of the sampling sequence is 2200, and the carrier frequency is 16 kHz. The start frequency of the 1st speech sensing sequence is 100 Hz, and the cutoff frequency is 150 Hz. The start frequency of the 2nd speech signal x₂(t) = sin[2π(100t + 100t²) is 100 Hz, and the cut-off frequency is 300 Hz. According to the above parameter settings, the detected translator speech sequence is obtained, as shown in Fig. 2, and the duration of the speech signal is 0~1s.

Figure 2.

Speech sequence

Taking the signal sequence of Fig. 2 as the research object, the designed method is compared with A system recognition based on sound signal, B recognition based on LM-BP neural network, and C system recognition method based on HMM to test the accuracy of different methods for speech recognition, and the comparison results are shown in Fig. 3.

Figure 3.

Comparison result

The four speech recognition methods were tested for 500 trials, respectively, and it can be seen from the figure that the method in this paper has the highest speech recognition accuracy, and with the increase in the number of trials, the speech recognition accuracy remains stable. The mean values of speech recognition for the system recognition based on acoustic signals, recognition based on LM-BP neural network, HMM-based lineage recognition method, and the method in this paper are 0.78, 0.81, 0.84, and 0.92, respectively.

The oral history materials exhibition’s designers used a range of digital devices, including giant screen projection, display screens, and touch devices, to enable visitors to immerse themselves in the richness of folk culture. In the digital art exhibition “Qingming Riverside Drawing 3.0”, in the part of “Long Scroll of the Sheng Shi”, the designers made thousands of static elements in the original painting into dynamic images, which showed the details of the painting completely and vividly in front of people. Combined with the oral history data of “Qingming Riverside Drawing”, the visualization design expands the information presentation of the exhibition.

The laser 3D projection calibration system mainly consists of a laser 3D projection system (referred to as a projector), a multi-parameter calibration system, and a laser tracker. The calibration wall in the multi-parameter calibration system fixes multiple reflective target heads and the target sphere of the laser tracker. The positions of each reflective target head and target sphere have been calibrated. At the same time, the multi-parameter projection frame is equipped with an iGPS receiver, whose vertex sphere seat and laser tracker target sphere seat have the same function and are also used to calibrate the center parameters of the projector.

In the whole giant screen projection, the projector projection range needs to cover the entire calibration wall, the projector outgoing laser on the calibration wall, reflective target head positioning, the laser tracking instrument on the calibration wall and the target seat on the projector positioning, through the solution of the relationship to get the projector coordinate system is located, i.e., the center of the projector, so as to achieve the optical center parameters of the laser 3D projector, and derive the calibration accuracy.

In this paper, matlab is used to simulate and analyze the accuracy and stability of this calibration model method. First of all, through the actual laser 3D projector working situation in the space arrangement of about 2m × 3m calibration wall, the calibration wall distribution of 10 × 10 target reflective head, in about 3m, 4.5m, 5m at the placement of the laser projector, the distribution of its distribution into the cubic situation, for 50 times simulation experiments.

In the simulation, station 1 out of 50 times, i.e., S1 position, is selected as the typical data enumeration. At the same time, the typical 6 target points out of 10×10 target points are presented. The theoretical target points for station S1 are shown in Table 1 for the position parameters before applying noise. Where H and V are the horizontal and pitch angles of the target point in the coordinate system of the laser 3D projector. The ^WP(x_PW,y_PW,z_PW) of target point P3 is (0.0001, 0.0005, 0.0001), respectively.

After setting the projector position state, this H and V angle can be solved to obtain, and a random measurement error with mean 0 and standard deviation σ = 1" obeying normal distribution is applied to the H and V corresponding to these 6 groups of target points as the true H and V angles observed by the projector. ^WP(x_PW,y_PW,z_PW) are the 3D coordinate values of the target points under the world coordinate system {W}. The position parameters of the theoretical target points at station S1 after applying noise are shown in Table 2. Also taking the target point P3 as an example, its position parameters ^WP(x_PW,y_PW,z_PW) after applying noise are (0.0001, 0.0005, 0.0001), which are consistent with those before applying noise.

In the simulation, the Newton iterative algorithm, Levenberg-Marquardt algorithm (L-M algorithm for short) and hybrid particle swarm algorithm are used to solve the observations respectively, and the results of the three solving algorithms are shown in Table 3. In terms of computational speed, the Newton iterative algorithm is the fastest, followed by the L-M algorithm and the BreedPSO algorithm. The observed values of the target point P1 in the Newton iterative algorithm are x_Pp/(mm) =-1365.6521, y_Pp/(mm) =996.852, and z_Pp/(mm) =3215.0231, respectively.

According to the simulation design, 50 simulations are carried out, and the hybridised particle swarm method is applied to solve the calibration parameters and calibration errors in the simulation to obtain the calibration error results for each simulation. The simulation results are in the range of 0.00458 to 0.00926, while the fused RMS results are 0.00315, so the method can eliminate the coarse error and reduce the random error. Among them, the fused coordinates are shown in Table 4, still taking the above six groups of target points as an example. The optical centre calibration parameter of station S1 is S1=[ −0.98691 -0.03596 0.019864 963.5219−0.05203 0.95238 -0.00091 -1024.5633−0.02494 -0.00186 -0.99865 3125.04260 0 0 1] .

Based on the arithmetic average of the 50 groups of optical centre parameters and the three coordinates of the calibration points under the respective coordinate system is the final optical centre parameter coordinate ^PV(x_Vp,y_Vp,z_Vp), which is the centre parameter coordinate of the giant screen projection scene.